Pipeline

Cleaning in progress. General steps below

1. Create dataset.

Set:

• samples (int)
• data_dir (str)

Can modify:

• Reward function
• outcome definition (one year vs three month vs time-to-follow-up)
• features
• actions
• state representation: <t, meld, previous meld change, candidate, donor>

python create_dataset.py

2. Create splits

Set:

• data_dir (str)

python create_splits.py

3. Train

Set:

• data_dir (str)
• alg (str)

Can modify:

• preprocessing
• encoders

python train.py

4. SOC

Set:

• data_dir (str)

python soc.py

5. Evaluate

Set:

• data_dir (str)
• rl model (str)

python test.py